Glass errors

[mcm]

Here's a sample of some odd artifacting that I've encountered trying to create glass bottle. The raw cube inside the bottle is just a backdrop to make the artifacting more visible.

I vaguely recalled that someone else had this same issue before, but I wasn't able to find how they solved it when I searched through the forums. The bottle was modeled in Silo and the only workable import into Blender 2.48a was a .dxf (.obj & .3ds didn't work). Using Blendigo/Indigo 1.1.18. Texture is just a standard glass-like Lasf9.nk that I've applied as a Specular with Transparent set, and very low gain of .05 and IOR of 1.54. I've checked normals, smoothed, unsmoothed and made solid, auto-smoothed, even edge-split and sub-surfed, and nothing seems to really help. The bottle is modeled as real bottle with a small, somewhat variable thickness throughout. It almost looks like there are some oddly-textured or doubled-up polys on the inside surface of the bottle. But I'm having a hard time determining for sure.

Could someone please jog my memory as to how to solve this? Thanks very much.

[Stur]

It looks like an overlapping faces issue. How thick is your glass ?

[mcm]

It is very, very thin (I'm not sure how to quantify it, otherwise). Before I exported the refined control mesh out of Silo, I double-checked to make sure that thickness was consistent enough so that the inside polys didn't go outside and vice-versa.

[PureSpider]

Be sure your mesh thickness is above the ray nudge disatance at every point!

[CTZn]

As PureSpider said, 0.2 mm is the strict minimal distance between one glass surface and any other type of surface, unless if you are explicitly using the precedence system (for materials using mediums).

But this kind of artifact may also have existed in older Indigo versions (not sure about 1.1.18) with one or the other scene tree building method (there are two), try forcing the other available one by changing bih_tri_threshold (as in the Indigo Technical pdf, page 16):

Given the final polycount for one mesh, a value below that count will enforce one method for the mesh, and vice-versa.

Dunno how it's called in blendigo 1.1.18... in blendigo at all in fact

[mcm]

Ah, thank you for the tip! Before I started redoing the mesh and completely changing the glass thickness, I did a quick experiment with the exponent for the nudge. That did the trick!

[pixie]

As PureSpider said, 0.2 mm is the strict minimal distance between one glass surface and any other type of surface, unless if you are explicitly using the precedence system (for materials using mediums).

But this kind of artifact may also have existed in older Indigo versions (not sure about 1.1.18) with one or the other scene tree building method (there are two), try forcing the other available one by changing bih_tri_threshold (as in the Indigo Technical pdf, page 16):

Given the final polycount for one mesh, a value below that count will enforce one method for the mesh, and vice-versa.

Dunno how it's called in blendigo 1.1.18... in blendigo at all in fact

This img the glass has no contact wih any other material yet the fireflies, oh the fireflies...

[CTZn]

Hey pixie, you are illustrating one of the weak points of bidir as a technique, wich is -approximatively- when rays have a path similar to:

Code: Select all

light > diffuse > specular > camera

Not "fireflies", but slow convergence in this case.

That's why Indigo comes with a panel of techniques that you can combine to suit your needs

[Stur]

Why such a thin glass mcm ? It's unrealistic, don't you think ?
You should let the ray nudge distance at its default value and make your glass a little bigger, in my humble opinion.